Claude Code v2.1.283
Released Sep 25, 2026 (19 days ago).
263 issues reference this build
211 still open
52 resolved (20%)
"References this build" means the issue body mentions running 2.1.283 (from the
bug-report template). It reflects report volume, not a proven defect count — an issue may mention a
version without being caused by it. First-party data from the anthropics/claude-code tracker.
Top issues reported on v2.1.283
Ranked by community engagement (reactions and comments).
- [FEATURE] customHeadersHelper: fetch ANTHROPIC_CUSTOM_HEADERS from a command (like otelHeadersHelper)
- [Bug] Classifier service unavailable or non-functional
- Messages meant for the user are sometimes emitted as hidden thinking before a tool call
- [BUG] VS Code extension: slash-command and `/goal` messages never collapse, so long /command arguments fill the chat panel
- [Bug] Weekly usage limit consumption rate increased ~3.6x after September 25 reset
- [BUG] Active Apple Max 20x subscription still recognized as Pro
- EVER SIGNLE TASK DONE BY CLAUDE CODE HAS TO BE REDONE ADN CORRECTED BY CODEX
- Remote Control: unarchived server-mode sessions are never re-dispatched to the running host — messages hang at "Sending…", then "Claude Code on <machine> is offline"
- [Bug] Server-side auto mode classifier returns "no verdict" errors; allow-listed commands blocked
- [BUG] 2.1.283: sandbox.excludedCommands never unsandboxes aws calls with --profile or --endpoint-url (Linux)
- USELSS PIECE OF SHIT CANT EVEN DOA. SIMPLE AUDIT
- [BUG] Turns hang ~6 minutes then "Connection dropped (ECONNRESET)", sudden onset 2026-09-28, across all sessions on one Mac
- [Bug] Anthropic API Error: Auto Mode Classifier No Verdict for Write Action
- Bash tool refuses read-only and worktree-local commands as 'too complex to verify' in git worktree isolation
- Subagent transcript view status line shows parent session model, not the subagent's model override
- [BUG] A changed MCP tool description never reaches a session that already loaded the tool, even after reconnect or --resume (deferred_tools_record is replayed)
- Agent isolation "remote" silently runs as a local worktree
- claude plugin eval: every Bash call fails on stock Ubuntu (apply-seccomp setgroups, #43454) and the sandbox settings workaround cannot be applied
- [BUG] /model: "Default (recommended)" shows your saved pick after restart
- [BUG] Sandboxed Bash tool blocks .git writes (refs/heads, index.lock) despite filesystem.allowWrite/excludedCommands overrides
- [BUG] Claude in Chrome: session tabs are created in the background and never activated, so they stay visibilityState "hidden" and screenshots of embedded apps/SPAs come back blank (macOS; same mechanism as #78533)
- Worktree isolation refuses every pwsh/powershell/cmd invocation with any argument; git-as-path-segment also refused inside interpreter text
- [BUG] VS Code: plugin MCP server silently dropped when a claude.ai connector has the same URL (CLI keeps the plugin)
- I'm ready to help generate a GitHub issue title for Claude Code. However, I don't see a bug report in your message. Could you please provide the bug report or issue description that you'd like me to create a title for?
- [Feature Request] Larger context window (2M–2.5M tokens) for long-lived codebases
- [FEATURE] VS Code extension: allow voice dictation in Remote - WSL windows (mic is available via WSLg)
- [Bug] Anthropic API Error: Auto Mode Classifier No Verdict for WebSearch
- [Bug] Anthropic API Error: Security Verifier Unavailable in Auto Mode
- [FEATURE] Agents view: show when each session last worked instead of how long it worked
- [BUG] Linux sandbox: filesystem.allowRead on a symlink inside a denyRead directory has no effect
Changelog
What's changed
- Added
x-claude-code-prompt-idto the gateway hint headers so LLM gateways can group the requests that serve one user prompt; opt in withCLAUDE_CODE_GATEWAY_HINT_HEADERS=1 - Added
availableModelsMatchmanaged setting: with"exact", anavailableModelsentry allows only the model version it names, so new releases stay blocked until listed - Added
deniedModelsmanaged setting to block specific models, even whenavailableModelsallows them - Added MCP tool, WebFetch and WebSearch outputs to the
tool.outputOpenTelemetry span event whenOTEL_LOG_TOOL_CONTENT=1 - Added
/doctor prompt-audit(also/checkup prompt-audit) to audit your CLAUDE.md files, skills, agents and commands for prompting patterns written for older models - Added click-to-expand for truncated messages from your other sessions in fullscreen mode
- Added
pathto--plugin-dirload-failure entries in the stream-jsonsystem/initplugin_errors, naming the directory that did not load - Added an opt-in
load_test_modeblock to the Claude apps gateway config: requests are built and signed but not sent upstream, and clients get a canned reply, so a deployment can be load tested - Added a
mantleupstream provider to the Claude apps gateway for Amazon Bedrock's Mantle endpoint - Fixed SDK sessions losing a deferred tool call or finished tool result when a turn ended early, a held approval prompt after a worker restart, and a non-streaming fallback's
result.usage - Fixed MCP progress notifications being discarded once a long-running tool call moved to the background; the background task now shows the latest progress
- Fixed stdio MCP servers being left running when the session ended while they were still starting
- Fixed a brief HTTP 404 from a stateless remote MCP server (for example a proxy mid-redeploy) leaving that server unusable for the rest of the session while still shown as connected
- Fixed MCP sign-in for a server with no valid URL failing with an opaque SDK error;
/mcpno longer offers Authenticate for such servers - Fixed the weekly Fable limit not appearing in
/usageand the VS Code usage meters when telemetry is disabled - Fixed
/modelaccepting Sonnet 4.6 or Sonnet 5 with[1m]when the id carried a date or-v1:0suffix, in the cases where the plain id was refused - Fixed
/modelpicker showing a hardcoded Haiku version and price whenANTHROPIC_DEFAULT_HAIKU_MODELpins a different model - Fixed dynamic workflows started during a model fallback running every agent on the fallback model instead of retrying the configured model
- Fixed
DISABLE_PROMPT_CACHING_HAIKUhaving no effect when Haiku is the session's main model - Fixed
claude plugin validatesaying Claude Code accepts a plugin or marketplace name it cannot install; such names inmarketplace.jsonnow fail validation - Fixed
claude plugin validatepassing plugins whoseoutputStyles,themes,monitors, orlspServerspaths are missing or point outside the plugin directory - Fixed
claude plugin detailsshowing 0 MCP servers for plugins that declare their servers inplugin.json - Fixed
claude plugin marketplace removenot saying which installed plugins it uninstalled with the marketplace; it now lists them - Fixed
claude plugin uninstallremoving the other of two installed plugins whose ids differ only in case, with its options and secrets, when the one named had noenabledPluginsentry at that scope - Fixed plugins that declare no version being silently restored at their source's newest commit, not the installed one, when their cached files were missing
- Fixed user-installed plugins and marketplaces failing to load with "cache-miss" after the home or config directory was moved, for example in bind-mounted devcontainers
- Fixed
installed_plugins.jsonshowing no plugins when it holds a record under an invalid plugin id; such a file loads again - Fixed
installed_plugins.jsonbeing rewritten, losing records, when it hol